Functional explanation
(from the Routledge Encyclopedia of Philosophy)
Explanations appealing to the functions of items are common
in everyday discourse and in science: we say that the heart pumps blood
because that is its function, and that the car fails to start because the
ignition is not functioning. Moreover, we distinguish the functions things
perform from other things they do: the heart makes a noise, but that is
not one of its functions. Philosophical discussions in this area attempt
to specify conditions under which it is appropriate to ascribe functions
to items and under which it is appropriate to appeal to those functions
in explanations. Difficulties arise because functions are normative: there
is some sense in which items ought to perform their functions; failure
to perform is a kind of error. Philosophical discussions investigate whether
and how this normativity can be understood in scientifically respectable
terms. This is important, because biological entities are among the most
characteristically functional items. This issue gives rise to differing
views as to what it is that functional explanations explain. One view is
that they explain how a containing system achieves some goal or effect.
Another is that functional explanations explain causally why the functional
item exists.
1. Historical context
Science and explanatory practice prior to the modern age
was based on the Aristotelian view that explanation is by appeal to causes,
and that the primary sense of ‘cause’ is final cause - telos or purpose
(see Aristotle §9). Activities of things which are parts of wholes
were understood as subserving the final causes of the wholes of which they
are parts. Their activities, insofar as understood, are their functions.
With the discovery of precise mathematically formulable laws of dynamics
for general application to all types of inanimate motion, appeal to purposes
falls not just out of service but into disrepute. Contemporary science
generally conceives nature as the realm of law-like regularity, devoid
of irreducible purpose and design. Explanation is still appeal to causes,
but these are now conceived in terms of regularity and mechanism. The advent
of genetics, the means by which Darwinian natural selection can account
for the origins of complex animate behaviour, gives further weight to this
conception (see Evolution, theory of).
Yet despite its apparent appeal to purposes and its lack
of overt appeal to natural laws, functional discourse and explanation has
persisted and been of great utility, especially in the biological sciences.
Contemporary discussions seek to provide for these practices a basis which
is naturalistic in the modern sense. Three interrelated problems beset
such attempts. First, not all effects are functional (the heart has the
mere effect, but not the function, of making sound); so functional must
be distinguished from ‘mere’ effects. Second, items may have functions
which they do not perform. A heart in fibrillation, and the eyes of a congenitally
blind person, still have their functions. Third is the specification of
what functional explanations explain.
2. Explaining the accomplishment of real goals
Some philosophers distinguish functional from mere effects
by taking functions to be contributions to real goals of containing systems,
whose accomplishment is partially explained by the resulting function ascription.
Ernest Nagel and Morton Beckner, among others, have offered naturalistic
criteria for such goals. Roughly, systems are goal-directed if and only
if they maintain some feature (possibly a developmental tendency) in relative
homeostasis through varying conditions, where that relative homeostasis
is achieved by the combined operation of multiple components of the system,
whose states are logically independent of one another but causally related.
The goal of the system may then be understood as the maintenance of the
homeostatic feature (or the state towards which the development tends).
Nagel himself recognized that such non-directed objects
as elastic solids and chemical systems in thermodynamic equilibrium are
not obviously ruled out by his proposed definition. Moreover, nothing in
this approach helps to solve the second, ‘malfunction’, problem, for, unless
more is said, a system which ceases to behave in a goal-directed manner
ceases to have a goal, and its subparts cease to have a function. Efforts
to solve this difficulty generally specify system types in terms of typical
members. Since tokens of a type can fail to be typical, token systems can
have typical goals they do not accomplish, and their components typical
functions they do not perform. However, attempts to distinguish typical
members have generally seemed either arbitrary or else to appeal, implicitly
or explicitly, to their proper functioning, rendering the account circular
for present purposes.
3. Explaining the presence of the functional
item
Following claims of sociologists such as Malinowski and
Merton that functional explanation is a distinctive method in sociology
(see Functionalism in social science), positivist philosophers, notably
Hempel and Nagel, attempted to assimilate functional explanation to the
deductive-nomological model thought to characterize all explanation. On
the deductive-nomological (D-N) model, explanation is a species of inference:
from the relevant laws of nature and further facts and relevant background
conditions, explananda can be deduced or, in the case of statistical laws,
induced (see Explanation §2). Function ascriptions are here taken
as equivalent to arguments in which the presence of the functionally characterized
item is inferred from a containing system’s having certain effects, and
from the need for an item of that sort to produce those effects.
It is rarely if ever the case, however, that the only
means by which a system can have a given effect is by means of a single
specific item. For example, the pumping of blood may be achieved by means
of an artificial device. For this reason, plausible function ascriptions
will fail as D-N explanations, because the inference from the effect to
the presence of the item in question will typically be unsound. This problem
has been generally regarded as fatal to any attempt to accommodate function
ascription to the D-N model.
The demand that explanations support valid or strong
statistical inferences has been generally attacked as too strict. Understanding
explanation more modestly, some philosophers continue to maintain that
legitimate function ascription explains the presence of functionally characterized
items. In a highly influential analysis, Larry Wright (1973) claims that
‘the function of I is F’ means
(a) I is there because it Fs;
(b) F is a result (or consequence) of I’s being there.
The first, etiological requirement is designed to
eliminate mere effects, but is clearly too weak to do the job. To cite
a well-known example, suppose a pebble has the effect of propping up a
rock in a stream. But for the presence of the rock, the pebble would be
washed away. Hence the pebble is where it is because it has the effect
of propping up the rock. On Wright’s analysis, then, the pebble has the
function of propping up the rock, which is surely a misapplication of the
concept of function. The account also seems under-inclusive, as the second
clause demands that the item actually have the functional effect, eliminating
mal- and non-functioning items.
The most plausible efforts to refine Wright’s basic etiological
framework attempt to solve these two problems at once, holding that the
etiologies which mark out functional effects are selectional, understood
roughly on the model of selectional explanation in evolutionary biology.
Importantly, selection always involves the relative replicative or reproductive
success of objects with a given feature as compared with competing objects
without that feature. The function of the feature is the effect it had
in virtue of which objects possessing it were selected. The ‘malfunction
problem’ is supposedly solved by the fact that the function of an item
is not identified with any of its actual or even possible effects, but
rather with selected-for effects of some predecessor items.
Whether selection history, even in the biological case,
actually provides a genuine etiology for particular items is disputed.
Robert Cummins (1975) and Elliott Sober (1984), for example, each deny
that it does, holding that the etiology of a biological trait is exclusively
a matter of the genetic plan inherited by the containing organism. Proponents
of the selection view counter that selection accounts for the increased
incidence of the item in the population, thereby explaining the fact that
the token system in issue possesses it.
Appeal to original selection etiology seems in any event
to give rise to further counterexamples, since the effects for which an
item has been selected may intuitively be non-functional or positively
disfunctional in subsequent generations of the containing systems, and
items may have effects in systems that are plausibly regarded as functional,
even where their etiologies are either non-selectional, or involve selection
for some other effects. Ruth Millikan (1984), whose influential selectional
theory of ‘proper functions’ is intended for use in a naturalized semantics
(see Semantics, teleological), has attempted to deflect such counter-examples
by claiming that her account is a theoretical definition, rather than an
analysis of our ordinary concept of ‘function’; failure of overlap with
our intuitive function ascriptions is beside the point. However, actual
selection histories are generally difficult if not impossible to determine,
and the boundaries of the theoretical concept, from an instrumental or
operational point of view, are no sharper than those of the intuitive.
Moreover, Millikan’s account is intended to capture not just biological
functions, but others as well, and it is not clear that the notions of
ancestry and replication it requires are applicable outside of the biological
context (or perhaps even within it).
Notice that the types of counterexample mentioned above
involve the temporal dimension (items can lose functions they once had,
and can gain functions after having been selected for other effects). Peter
Godfrey-Smith (1994) and Paul Griffiths (1992, 1993), each concerned particularly
with biological function, have suggested that these counterexamples can
be avoided if only recent or proximal selection history is regarded as
relevant to function ascription. The viability of such accounts depends
upon whether the notions of ‘recent’ or ‘proximal’ history can be made
precise. This is turn depends upon empirical assumptions about likely mutation
rates and the incidence and nature of competitors, as well as the assumption
that evolution is efficient in the sense that, if an item is not providing
some fitness benefit to an organism, then the incidence of that item in
the population will decline. Though not implausible, such assumptions are
far from clearly warranted. Any attempt similarly to refine a selectional
account of function for general application beyond the context of biology
would involve a more dubious generalization of these assumptions. Finally,
some biological items with selectional etiologies, notably, so-called ‘segregation
distorters’, provide no benefit to containing systems, and are even positively
harmful. Unless we wish, counter-intuitively, to regard these effects as
functional, some non-arbitrary means of ruling them out of the analysis
is needed.
4. Explaining interesting capacities
Because of difficulties of these kinds, some philosophers
deny that the distinction between functional and mere effects is ultimately
real and that function ascriptions do or can operate as explanations of
the presence of the functional item. Robert Cummins has offered a well-known
account along these lines, on which ascriptions of functional properties
to items are always relative to an analytical account of some capacity
of some containing system, that is, to an account which breaks down the
net capacity into sub-processes which, taken together, are sufficient to
explain the capacity. On this view, any capacity of a containing system
which results from contributions of its parts is subject to functional
analysis, and such parts could conceivably be ascribed innumerable distinct,
even incompatible functions (though because of the relativization of function
ascription to analysis, no inconsistency results). Cummins’ analysis is
thoroughly instrumental, holding that our explanatory interests determine
which effects we single out in the sciences for this kind of analysis,
hence our ascriptions of functions to particular items, and also which
tokens we take as typical, hence what we treat as malfunctional.
There is little doubt that Cummins’ functions are naturalistically
respectable, so long as the particular analyses offered are themselves
naturalistic, and that they capture much of the actual practice of functional
explanation. However, function ascription along Cummins’ lines seems inadequate
to account for those cases in which selectional or other etiological accounts
adverting to function can be given, notably in the case of artifacts. Moreover,
Cummins’ functions are plainly inadequate for service in broader philosophical
projects, such as naturalized semantics and epistemology, which require
a robustly realist conception of the normativity of function.
See also: Causation;Teleology; Technology, philosophy
of
RICHARD N. MANNING
References and further reading
Beckner, M. (1959) The Biological Way of Thought,
New York: Columbia University Press.(See chapters 6 and 7. Moderately technical
accounts of functions as contributions to goals and of goal-directed systems,
in the biological context.)
Bigelow, J. and Pargetter, R. (1987) ‘Functions’, Journal
of Philosophy 84: 181-97. (Functions as contributions to goals).
Boorse, C. (1976) ‘Wright on Functions’, Philosophical
Review 85: 70-86.(Counterexamples to the etiological theory.)
Cummins, R. (1975) ‘Functional Analysis’, Journal
of Philosophy 72: 741-65. (Detailed critique of positivist and etiological
theories of function; presentation of instrumental account.)
Davies, P.S. (1994) ‘Troubles for Direct Proper Functions’,
Nous
XXVII: 363-81.(Somewhat technical challenge to the applicability of Millikan’s
theory of proper functions to biological cases.)
Godfrey-Smith, P. (1993) ‘Functions: Consensus Without
Unity’, Pacific Philosophical Quarterly 74: 196-208.(Highly readable
presentation of the etiological and instrumental accounts as distinct and
equally important.) Godfrey-Smith, P. (1994) ‘A Modern History Theory of
Functions’, Nous 27: 344-62.(Attempts to accommodate loss and gain
of functions within the selection etiology framework, along with a discussion
of segregation distortion.)
Griffiths, P. (1992) ‘Adaptive Explanation and the Concept
of a Vestige’, in P. Griffiths (ed.) Trees of Life, Dordrecht: Kluwer,
111-31.(Somewhat technical attempt to accommodate loss and gain of function
within the selection etiology framework.)
Griffiths, P. (1993) ‘Functional Analysis and Proper
Functions’, British Journal of Philosophy of Science 44: 409-22.(As
above, with a discussion of segregation distortion and an extension to
artifacts.)
Hempel, C.G. (1965) ‘The Logic of Functional Analysis’,
in Aspects of Scientific Explanation, New York: Free Press.(Classic
discussion of functional explanation as a species of D-N explanation.)
Kitcher, P. (1988) ‘Function and Design’, Midwest
Studies in Philosophy 18: 379-97.(Attempts to unify function under
the concept of design; moderately difficult.)
Manning, R.N. (1997) ‘Biological Function, Selection,
and Reduction’, British Journal for the Philosophy of Science 48:
69-82.(Rejecting reductive status of selectional accounts of function;
moderate difficulty.)
Millikan, R.G. (1984) Language, Thought, and Other
Biological Categories, Cambridge, MA: MIT Press. (Influential theory
of ‘proper functions’ for use in naturalized semantics; technical but rewarding.)
Millikan, R.G. (1993) White Queen Psychology and other
Essays, Cambridge, MA: MIT Press.(Further clarification, defence, and
application of ‘proper function’.)
Nagel, E. (1953) The Structure of Science, New
York: Harcourt Brace.(See chapter 12. A typical account, in a positivist
framework, of goal-directedness for systems; somewhat technical.)
Neander, K. (1988) ‘What Does Natural Selection Explain?
Correction to Sober’, Philosophy of Science 55: 422-6.(Nontechnical
defence of the view that selection explains the presence of selected for
items.)
Neander, K. (1991a) ‘The Teleological Notion of "Function"’,
Australasian
Journal of Philosophy 69: 454-68.(A selectional account of function;
moderate difficulty.)
Neander, K. (1991b) ‘Functions as Selected Effects: The
Conceptual Analyst’s Defence’, Philosophy of Science 58: 168-84.(Rejecting
Millikan’s defence of ‘proper function’ as a theoretical definition.)
Sober, E. (1984) The Nature of Selection, Cambridge,
MA: MIT Press.(See chapter 5. Argues that selection does not explain the
presence of biological items; moderately difficult.)
Sober, E (1985) ‘Panglossian Functionalism and the Philosophy
of Mind’, Synthèse 64: 165-93. (Nontechnical discussion of
functionalism in biology, and its possible use in accounting for mental
contents.)
Wright, L. (1973) ‘Functions’, Philosophical Review
82: 139-68.(An etiological theory of function; highly readable.)
Wright, L. (1976) Teleological Explanations, Berkeley,
CA: University of California Press.(Booklength version of the above.)
from the
Routledge Encyclopedia of Philosophy, Version 1.0, London:
Routledge
GO
BACK TO COURSE HOME-PAGE